Multiple scattering reflection models theory, implementation and experimental analysis
Abstract/Contents
- Abstract
- Physically-based rendering (PBR) techniques have been widely used in photo-realistic and stylized image generation across film and real-time graphics. Along with physically based lighting models, microfacet-based reflection models developed using physical principles of surface reflectance form the foundation of PBR. While microfacet-based reflection models can represent light reflections on a wide range of rough surfaces, a major limitation of these models is that they only account for the single scattering of light, leading to energy loss and image darkening. Before PBR, artists used ad-hoc diffuse additions to compensate for this energy loss, but these types of fixes are difficult to use in PBR where any excess energy can cause images to blow out. Therefore, the broad adoption of PBR made this limitation even more pronounced for the application of microfacet BRDF models. In this thesis, we present two approaches to represent surface reflectance with multiple scattering effects. The first approach is the first closed form multiple scattering microfacet BRDF that leverages Zipin's classic paper which proves the number of reflections inside a specular v-groove structure is bounded. This solution improves upon the prior state-of-the-art multiple scattering BRDF by 10𝑥+ in terms of path-traced rendering convergence and performance. The second approach leverages advances in probabilistic distribution modeling using deep neural networks to develop a multiple scattering BRDF using Real-NVP. This is the first work to demonstrate the potential of normalizing flows for BRDF modeling and sample generation. In addition to presenting the theory and the implementation of these two models inside the PBR renderer, PBRT, we also present the first comprehensive experimental analysis comparing three state-of-the-art multiple scattering reflection models with the three most widely used single scattering reflection models (the GGX Smith, GGX V-groove and Cook Torrance BRDFs) using the MERL material database, a dataset with 100 measured isotropic materials with varying albedo, roughness and Fresnel index of refraction.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2022; ©2022 |
| Publication date | 2022; 2022 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Xie, Feng |
|---|---|
| Degree supervisor | Hanrahan, P. M. (Patrick Matthew) |
| Thesis advisor | Hanrahan, P. M. (Patrick Matthew) |
| Thesis advisor | James, Doug L |
| Thesis advisor | Marschner, Steve |
| Degree committee member | James, Doug L |
| Degree committee member | Marschner, Steve |
| Associated with | Stanford University, Computer Science Department |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Feng Xie. |
|---|---|
| Note | Submitted to the Computer Science Department. |
| Thesis | Thesis Ph.D. Stanford University 2022. |
| Location | https://purl.stanford.edu/fb632hc0479 |
Access conditions
- Copyright
- © 2022 by Feng Xie
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...